Bram Stein

It’s here! Go buy the Webfont Handbook and learn about rendering, licensing & optimising your webfont loading! abookapart.com/products/webfo…

For the first time since 2007, you can buy fonts directly from my website again. Also, live, editable font previews. At last! marksimonson.com/fonts

My latest typeface release is Dreamboat, based on the classic early twentieth century "bold script" style of lettering, still seen on logos for baseball teams, car companies, beer and soft drink brands. More info: marksimonson.com/notebook/view/… marksimonson.com/fonts/view/dre…

@Lorp @justvanrossum @NickSherman @davelab6 @googlefonts This one: httparchive.org You can access the data with BigQuery: github.com/HTTPArchive/ht… The code that extracts data from fonts is here: github.com/HTTPArchive/wp… (you can query for counts.num_glyphs).

@bram_stein @justvanrossum @NickSherman @davelab6 @googlefonts Yes, typical numGlyphs for fonts on the web is what I was wondering. Which HTTP Archive are you talking about?

Variable font usage has grown to an impressive 29% (up from 13%). About 97% of variable font usage is from @GoogleFonts; only 3% from self-hosting and other services. Most people use variable fonts for performance and not for their typographic potential. #variable-fonts" target="_blank" rel="nofollow noopener">almanac.httparchive.org/en/2022/fonts#…

@Lorp @justvanrossum @NickSherman @davelab6 @googlefonts Do you mean typical numGlyphs for fonts on the web? If so, that data is available in the HTTP Archive.

@justvanrossum @NickSherman @davelab6 @bram_stein @googlefonts It would be interesting to know typical numGlyphs, if what Nick observes about smaller fonts compressing better with CFF2 than TTF is true. Webfonts of ~1500 glyphs are usually not the best choice.

@NickSherman @davelab6 @justvanrossum @googlefonts There shouldn't be a big difference (apart from some metadata in the CFF table). Perhaps the difference is in the other tables? The article compares only the outline tables.

@bram_stein @davelab6 @justvanrossum @googlefonts The other tricky thing is how all the moving parts for any given project can swing results. I even see surprising differences in file sizes for fonts exported from the same sources, to the same target format, but using different tools (Fontmake, AFDKO, etc).

@NickSherman @davelab6 @justvanrossum @googlefonts Me too! There are 189 CFF2 fonts used on the web compared to 9 million glyf and nearly a million CFF fonts, so not a fair comparison. Based on the spec I don't expect a huge difference: CFF2 will be slightly smaller, but not a strong reason for picking one format over the other

@bram_stein @davelab6 @justvanrossum @googlefonts I’d be curious to see comparisons between TTF and CFF2 equivalents, with WOFF2 compression, across a varied range of different type families. I’m happy to share data for my own fonts, but they are decidedly less extensive than things like Roboto, so the results will surely vary.

@NickSherman @davelab6 @justvanrossum @googlefonts One small correction the data shows CFF (not CFF2) vs. glyf sizes. Over the entire dataset the difference is negligible (and should not be a reason for picking one over another). There insufficient data on CFF2 for a conclusion (though I doubt there will be a large difference).

@davelab6 @justvanrossum @bram_stein @googlefonts TTFs are much bigger going in, so they benefit more from WOFF2 compression, percentage-wise. But as this article explains (if I’m reading it correctly), the CFF2 WOFF2s are still often smaller than TTF WOFFs equivalents, especially for smaller fonts.

@justvanrossum @NickSherman @googlefonts Only about 25% of the fonts on Adobe Fonts are CFF based (source: the WOFF2 evaluation report). It's probably a combination of all the things mentioned. I agree that the statement in the article is too simplistic.

@NickSherman @bram_stein @googlefonts “Type designers can choose the outline format they prefer” — not if you publish with Google Fonts. Maybe the apparent bias towards TTF has something to do with this.

@simoncozens @Lorp Textures with alpha channels!

@PixelAmbacht Yes, it is official. Doom runs on variable COLRv1 fonts.

@simoncozens I think the hidden surface removal can be baked into the animation axis using intermediate regions, but it is a little awkward. Tying compositing modes to variable axes would simplify a lot!

@simoncozens Yea, I was hoping for something that works as you transform (i.e. a front-facing surface becomes a back-facing surface). For example, if I wanted to spin this glyph around.

@simoncozens Nice! I've been working on converting arbitrary 3D models into variable/colr by projecting them into the 2D plane using outline based triangles. Haven't quite figured out hidden surface removal yet, but getting there.

@simoncozens Ah yes!

@simoncozens Hoping :) A 3D mesh baked into a variable font?

@simoncozens Ohhh, if that's what I think it is, we should talk!

@KhaledGhetas I agree. With only a handful of CFF2 fonts in existence, and the only tangible benefit of CFF2 being cubic beziers, wouldn't it make more sense to stop investing in CFF2, mark it as deprecated and move on?

@bram_stein I totally agree, but a new glyf version is not going to magically fix any of the issues preventing CFF2 adoption.

Slightly over 99.99% of variable fonts use TrueType glyph outlines. Not a good look for the CFF2 format six years after it was standardized. Perhaps time to deprecate it? #variable-fonts" target="_blank" rel="nofollow noopener">almanac.httparchive.org/en/2022/fonts#…

@KhaledGhetas I think that's one of the reasons: lack of tooling, implementations, etc. With usage this low, I wonder if resources wouldn't be better spent on other things (like adding cubic bezier support to glyf).

@bram_stein Isn’t this because Windows does not support CFF2?

@heydonworks @zachleat 👋Hi Heydon!